Wikipedia turns 25 this year facing an existential crisis that most people haven't noticed yet. AI companies scraped it for training data while search engines increasingly bypass it with AI-generated summaries. Meanwhile, the local journalism that Wikipedia relies on for sourcing is collapsing. We're watching a tragedy of the commons play out in real-time—AI companies extracting value from Wikipedia faster than the ecosystem can regenerate.
The irony is brutal.
Every major AI model was trained on Wikipedia. OpenAI's GPT series, Google's Gemini, Anthropic's Claude, Meta's Llama—they all scraped Wikipedia for free, high-quality, structured knowledge. It was the perfect training dataset: factual, comprehensive, multilingual, and continuously updated by volunteers.
Now those same AI products are replacing Wikipedia in search results.
When you ask ChatGPT a factual question, you don't need to click through to Wikipedia. The model gives you an answer synthesized from its training data—much of which came from Wikipedia. Google's AI Overviews work the same way: they show an AI-generated summary at the top of search results, reducing clicks to the underlying sources.
This is crushing Wikipedia's visibility and traffic. Fewer search result clicks mean fewer casual readers discovering articles. Fewer readers mean fewer potential editors. Fewer editors mean declining article quality and update frequency. The flywheel that kept Wikipedia thriving for two decades is starting to spin backward.
The second threat is harder to see but potentially more damaging: the collapse of local journalism.
Wikipedia doesn't generate original reporting. It synthesizes information from reliable sources—primarily newspapers, academic journals, and reputable news organizations. When those sources disappear, Wikipedia has nothing to cite.
Local newspapers have been dying for years. The ones that survive are cutting staff and reducing coverage. AI tools like ChatGPT and Perplexity are accelerating the decline by letting users get information summaries without visiting news sites or seeing ads. The economic model supporting journalism is collapsing.
Without journalism, there are no sources for Wikipedia articles about current events, local politics, or community issues. Wikipedia can document what happened, but only if someone else reported it first. When that reporting stops, the gaps in Wikipedia's coverage will grow.
The combination is devastating. AI trained on Wikipedia is reducing Wikipedia's reach while simultaneously destroying the journalism ecosystem Wikipedia depends on.
It's like strip-mining the forest you live in.
Wikipedia survived the decline of print encyclopedias by being free, comprehensive, and community-driven. Encyclopedia Britannica couldn't compete with millions of volunteer editors updating articles in real-time. But I'm not sure Wikipedia survives being used as free training data for trillion-dollar companies whose products directly compete with it.
The financial pressure is also mounting. Wikipedia is run by the Wikimedia Foundation, a nonprofit funded primarily by individual donations. The annual fundraising campaigns work because people value Wikipedia and understand it needs support. But if AI answers replace Wikipedia in daily use, will people still donate?
Younger users increasingly get information from AI chatbots, not Wikipedia. If an entire generation grows up never visiting Wikipedia, the donation base dries up.
Wikimedia Foundation had about $180 million in revenue last year, mostly from donations. That's enough to keep servers running and support a small staff, but it doesn't create the kind of financial cushion that could weather a major decline in visibility or donor engagement.
There are potential solutions, but none are easy.
Wikipedia could restrict AI companies from scraping its content. But Wikipedia's entire philosophy is built on free access and open licensing. Blocking AI training would require abandoning core principles and would be technically difficult to enforce.
AI companies could pay licensing fees to Wikipedia. OpenAI signed content deals with publishers like The Atlantic and Axel Springer. In theory, they could pay Wikimedia Foundation for training access. But Wikipedia's Creative Commons license already permits commercial use, so there's no legal obligation to negotiate.
Search engines could prioritize linking to Wikipedia instead of showing AI-generated summaries. But Google, Microsoft, and others are heavily invested in AI as the future of search. They're not going to voluntarily reduce their AI features to drive traffic to external sites.
Wikipedia could build its own AI tools and use them to enhance the editing experience or generate draft articles for human review. This might help with editor recruitment and retention. But it doesn't solve the traffic and visibility problem.
The most realistic path forward is that Wikipedia becomes less visible but remains a critical infrastructure that AI companies depend on indirectly. Instead of users reading Wikipedia articles, they'll read AI-generated summaries based on Wikipedia content. Wikipedia becomes plumbing—essential but invisible.
That might be sustainable if donations hold up and the editor community remains active. But it's a much more fragile equilibrium than Wikipedia has operated under for the past 25 years.
The deeper issue is that we're seeing a pattern repeat across the internet. Platforms and companies built enormous value by aggregating user-generated content and open knowledge. Then they used that content to train AI models that compete with the original sources. And the original sources—Wikipedia, news sites, niche forums—start to wither.
It's not malicious. It's just economics. AI companies optimize for user engagement and revenue. If giving direct answers reduces clicks to external sites, that's not their problem to solve.
But it becomes everyone's problem when the knowledge ecosystem that made AI possible in the first place starts to collapse.
Wikipedia has survived a lot of existential threats. Vandalism, misinformation, funding challenges, editor burnout, competition from other platforms. It's been resilient because the community is committed and the model is flexible.
But this threat is different. It's not something Wikipedia can solve by improving moderation or recruiting more editors. It's a structural shift in how people access information, driven by AI products that used Wikipedia to train but now compete with it for attention.
Wikipedia at 25 should be celebrating a triumph of crowdsourced knowledge. Instead, it's facing a future where it might become invisible infrastructure for AI companies worth trillions, receiving neither recognition nor compensation.
The technology is impressive. The irony is tragic. And the outcome is far from certain.
